DSAIL-TreeVision: A software tool for extracting tree biophysical parameters from stereoscopic images
Cedric Kiplimo,
Collins Emasi Epege,
Ciira wa Maina,
Billy Okal
Affiliations
Cedric Kiplimo
Department of Electrical and Electronic Engineering, Dedan Kimathi University of Technology, P.O. Box – PRIVATE BAG – 10143, Dedan Kimathi, Nyeri, Kenya; Centre for Data Science and Artificial Intelligence (DSAIL), Dedan Kimathi University of Technology, P.O. Box – PRIVATE BAG – 10143, Dedan Kimathi, Nyeri, Kenya; Corresponding author at: Centre for Data Science and Artificial Intelligence (DSAIL), Dedan Kimathi University of Technology, P.O. Box – PRIVATE BAG – 10143, Dedan Kimathi, Nyeri, Kenya.
Collins Emasi Epege
Department of Electrical and Electronic Engineering, Dedan Kimathi University of Technology, P.O. Box – PRIVATE BAG – 10143, Dedan Kimathi, Nyeri, Kenya; Centre for Data Science and Artificial Intelligence (DSAIL), Dedan Kimathi University of Technology, P.O. Box – PRIVATE BAG – 10143, Dedan Kimathi, Nyeri, Kenya
Ciira wa Maina
Department of Electrical and Electronic Engineering, Dedan Kimathi University of Technology, P.O. Box – PRIVATE BAG – 10143, Dedan Kimathi, Nyeri, Kenya; Centre for Data Science and Artificial Intelligence (DSAIL), Dedan Kimathi University of Technology, P.O. Box – PRIVATE BAG – 10143, Dedan Kimathi, Nyeri, Kenya
Billy Okal
NVIDIA Corporation, 2701 San Tomas Expressway, Santa Clara, CA 95050, United States
Reliable forest data is crucial for policy and investment decisions, hence the need to monitor forests. Tree harvesting policies, informed by tree attribute data, can help prevent excessive logging. The use of stereoscopic vision for estimating tree attributes is an active field of research. DSAIL-TreeVision is open-source software for extracting tree diameter at breast height, crown diameter, and height from stereoscopic images. It is written in Python and comprises three modules for capturing and storing images, calibrating single and stereo cameras, and extracting the tree attributes. While other open-source software implementing stereoscopic vision for other applications exists, no software for estimating these tree attributes exists. DSAIL-TreeVision was developed to fill this gap and has been tested in a real forest setting and on the DIML/CVL RGB-D dataset.
